From Text to Storyboard: Vidu Q4 Reference-to-Video Concept Demo

This entry demonstrates the reference-to-video workflow of vidu-q4-preview: first use two reference images to lock down the character and scene, then use a prompt to refer to the materials in the order of @Image1 and @Image2, driving a short shot. The topic focuses on the storyboard concept of "from text to picture"—breaking down a script description into reusable character design and scene design images, then handing them to reference mode to generate a continuous shot with native audio, avoiding repeated rewriting of the character's appearance. In terms of composition, the base retains the non-media parameters of the pipeline schema: duration is set to 6 seconds, 720p, 16:9, and audio remains enabled by default so that the same workflow includes dialogue and sound effects. Assets handles the two reference images: a half-body character design and a scene atmosphere image, bound to reference_image_urls[0] and reference_image_urls[1] respectively. Referencing the reference images in order in the prompt is a usage described in the pipeline schema documentation. Use cases: independent creators and small studios doing shot previews before formal production; serialized short films that require consistency of characters and style across shots; and scenarios where the same set of reference images is used to repeatedly compare expressions, camera angles, or opening styles. Note that this version is Preview, capabilities are continuously iterated, the maximum duration per generation is 16 seconds, and long shots need to be spliced in segments.